Abstract: In view of the rough selection of hysteresis time in the existing statistical model of piezometric level in the earth-rock dam,it cannot reasonably reflect the cumulative hysteresis effect of reservoir water level and rainfall. This paper firstly introduces the hysteresis effect function to reflect the influence of continuous change of reservoir water level and rainfall on seepage. Then particle swarm optimization is used to identify the hysteresis days and influence days of reservoir water level and rainfall. Finally,a statistical model for piezometric level of the earth-rock dam considering hysteresis effect is established,and the effectiveness of the new model is demonstrated by combining nearly20 years seepage monitoring data from the composite geomembrage embankment earth-rock dam of Wangfuzhou Hydraulic Project. The results show that compared with the traditional model,when the water level of piezometric fluctuates greatly,the statistical model with hysteresis function has a better fitting effect. The lag days influenced by the water level of GY5 section reservoir increase with the increase in the distance between the piezometric and the water surface. The monitoring data of GY5-2 are divided into 6 time periods and calculated. It is found that the lag days affected by reservoir water level basically remained stable at 16d,indicating that the working behavior of composite geomembrane in GY5 section maintained well.